Skip to content

Dataset · zlab-princeton

600K examples for visual reasoning RL

zlab-princeton/Vero-600k

A curated collection of 59 English datasets for training VLMs in logic, STEM, charts, and OCR.

What's inside

The dataset contains 600,000 curated image-text pairs in English for reinforcement learning (RL). The data is gathered from 59 different sources and covers 6 broad categories, including STEM tasks, charts and OCR, as well as spatial reasoning and actions.

What you can do with it

  • RL fine-tuning of multimodal models for logical reasoning skills
  • Improving VLM accuracy in chart recognition, text (OCR), and solving scientific tasks
Tasks
Image text → text
Languages
English
Size
100K–1M rows
Downloads this month
50.5K
Likes
47

How to load it

from datasets import load_dataset

ds = load_dataset("zlab-princeton/Vero-600k")